The role of abnormal DNA helicase function in the deregulation of DNA repair and genomic stability, and in human cancer development, is well documented. The functional interaction between the DNA helicase BRIP1 and the breast/ovarian cancer susceptibility gene BRCA1 makes common variants in BRIP1 good candidates for low to moderate penetrance susceptibility to both breast and ovarian cancer.
In this study, we evaluated the association between SNPs that efficiently tag the common variation in the BRIP1
gene and the risks of breast and ovarian cancer using a case-control study design. The tSNPs that we tested were not chosen because they were of any known functional significance. tSNP selection was based on the latest Hapmap data (Phase II release # 20) in CEPH DNA samples, which are of North Western European ancestry. The tSNPs selected using these data provide good power to capture all common variation 
. Therefore, we are confident that the set of tSNPs we chose adequately tag the known and unknown common variants within the gene.
Previous studies have looked at the association between a handful of common functional BRIP1
polymorphisms and breast cancer risk. The Ser919Pro variant was found to increase breast cancer risk in families using a ‘kin-cohort’ study design 
. However, subsequent studies failed to confirm this finding 
. Ser919Pro (MAF
0.45) was tagged by rs1557720 (rp2
0.97). We found no evidence of association between this SNP and breast cancer risk (P
0.39). Our study has 97% power to detect an allele with this frequency with a type I error rate of 0.0001, even if the true relative risk was 1.3.
Another study has indicated that the rare variant Arg173Cys (tSNP rs4988345) impairs protein translocation to the nucleus and might modify breast cancer susceptibility 
. Since the minor allele for this variant is very rare (MAF
0.008) in CEPH samples, we have very limited power to confirm or refute this association using the case-control population in the current study.
We found no evidence of association with breast cancer risk for any of the tSNPs analysed in our study. The staged design provides at least 85% power at a Type I error rate of 10−3
to detect an allele that explains 0.75% of the excess familial risk of breast cancer and has been tagged with r2
0.8 – e.g.a co-dominant allele with a frequency of 0.1 that confers a relative risk of 1.27. Therefore, it is unlikely that common variants in BRIP1
contribute significantly to breast cancer risk. However, we cannot exclude the possibility that the alleles we investigated are associated with smaller risks. Power to detect alleles explaining 0.5% of the excess familial risk is approximately 65%, and power to detect rarer susceptibility variants that are weakly correlated with the polymorphisms we examined is low. Furthermore some known common variants were poorly tagged, because of tSNP assay failure. Again power to detect association with these SNPs is limited.
To our knowledge, this is the first study to evaluate the association between common variants in BRIP1
and the risks of epithelial ovarian cancer. We found some evidence of association with disease risk for two of the 12 tSNPs tested- rs2191249 and rs4988344. However, these associations were only of borderline significance, and so need to be interpreted with some caution. Despite the large sample size, neither association is highly significant (P
0.047 and 0.02 respectively) and the P-values have not been adjusted for multiple hypothesis testing. As the tSNPs are correlated, the test statistics are not independent, and standard methods for adjusting for multiple testing, such as the Bonferroni correction, are too conservative. Therefore we used a simulation to determine an empirical P-value for the most significant result (P-trend
0.02 for rs4988344). In this analysis, we randomly shuffled the case-control status among individuals multiple times, and estimated how frequently a P-value<0.02 was obtained from the randomly permutated data. This method also accounts for the testing of multiple genetic models with each SNP. In 1,000 permutations a P<0.02 was observed on 357 occasions, giving the most significant P-value corrected for multiple testing of 0.36. Thus it is likely that the positive result is a chance finding.
Disease heterogeneity could also lead to false positive reporting or mask the presence of true associations. When we stratified cases according to histological sub-type we found that the strength of association with ovarian cancer risks improved for rs4988344 and rs2378908 when cases of serous histology only were considered (P trend
0.008 and 0.015 respectively). Once again, caution is required when interpreting these data. There is an inevitable loss of statistical power when stratifying cases into clinical sub-types, and after adjusting for multiple testing, the most significant of these associations (P-trend
0.008 for rs4988344) was P
Another explanation for a spurious association could be hidden population stratification. This occurs when allele frequencies differ between population sub-groups and cases and controls are drawn differentially from those sub-groups. The three ovarian cancer case-control studies used in the present analysis were from the UK, Denmark and USA. However, all analyses were restricted to subjects of the same ethnic origin (white, Western European) and so population stratification is unlikely to be explanation for erroneous associations. Even if stratification were present, it is unlikely that the same degree of stratification would occur in all three studies.
In combination, the three ovarian cancer studies has more than 80% power to detect a common allele that explains 0.75% of the excess familial risk at a type I error rate of 10−3 (for example an allele with frequency 0.2 that confers a relative risk of 1.3), and more than 55% power to detect a common allele that explains 0.5% of the excess familial risk.
Assuming the genetic associations we identified are real, they may either be due to a direct causative effect of the SNPs tested, or because these tSNPs are in linkage disequilibrium and serve as markers for the real determinant of a disease. rs2191249 and rs4988344 are both intronic and neither is strongly correlated with other known SNPs that are more likely to have functional role. The bioinformatics tool PupaSNP (http://pupanp.bioinfo.cnio.es
) suggests neither variant has a functional effect or dramatically alters the structure of BRIP1
. Thus, it seems likely that any true causal variant(s) will be in linkage disequilibrium with rs2191249 or rs4988344. All the known common coding SNPs, 3′UTR SNPs and 5′UTR SNPs were tagged by our selected panel of tSNPs with r2
>0.95 and were not associated with disease. However, we cannot exclude the possibility that unidentified variants exist in the promoter or regulatory region or intron-exon boundaries, which affect the transcription of BRIP1,
and are tagged by the two tSNPs for which we find association.
In conclusion, we have genotyped 12 tSNPs that tag the common variants in BRIP1 in breast and ovarian cancer case control series. We found no association with breast cancer risk for any tSNP; but we found evidence of borderline significant associations with invasive ovarian cancer risk for two tSNPs of unknown related function to BRIP1. The observed associations with ovarian cancer risk warrant further evulation in independent case-control studies.